Update TensorBoard training metrics

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 598315.063954628,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 76927.1618999912,
-      "required_pages_per_second_upper_estimate": 0.005321999555848895
+      "required_token_positions_per_second": 76961.01291187802,
+      "required_pages_per_second_upper_estimate": 0.005324341447396394
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 8533.957213384612,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9615.8952374989,
-      "required_pages_per_second_upper_estimate": 0.004025974533214191
+      "required_token_positions_per_second": 9620.126613984752,
+      "required_pages_per_second_upper_estimate": 0.0040277461221876285
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 380144.35761751566,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19231.7904749978,
-      "required_pages_per_second_upper_estimate": 0.004005474779521079
+      "required_token_positions_per_second": 19240.253227969504,
+      "required_pages_per_second_upper_estimate": 0.004007237347787281
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 43742.16556775435,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28847.6857124967,
-      "required_pages_per_second_upper_estimate": 0.0017293747105985482
+      "required_token_positions_per_second": 28860.379841954258,
+      "required_pages_per_second_upper_estimate": 0.0017301357042767156
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 22018.86276523721,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28847.6857124967,
-      "required_pages_per_second_upper_estimate": 0.0022427806765870596
+      "required_token_positions_per_second": 28860.379841954258,
+      "required_pages_per_second_upper_estimate": 0.0022437675893168104
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 910315.184299418,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 23,
-      "required_token_positions_per_second": 28847.6857124967,
-      "required_pages_per_second_upper_estimate": 0.37002713808823257
+      "required_token_positions_per_second": 28860.379841954258,
+      "required_pages_per_second_upper_estimate": 0.37018996475101984
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,14 +158,14 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.05980415519528748,
-    "producer_p95_seconds": 0.1412130310000066,
-    "gpu_update_seconds": 0.9584131037597584,
-    "producer_batches_per_second": 16.721246153123474,
-    "gpu_batches_per_second": 1.043391410318891,
-    "producer_headroom": 16.02586142434599,
+    "producer_batch_seconds": 0.06026643610157123,
+    "producer_p95_seconds": 0.15971326699946076,
+    "gpu_update_seconds": 0.9579915493630536,
+    "producer_batches_per_second": 16.592983834561416,
+    "gpu_batches_per_second": 1.043850544052165,
+    "producer_headroom": 15.895938292227594,
     "sustainable": true,
-    "buffer_seconds": 60.38002553686478,
+    "buffer_seconds": 60.35346760987238,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 82,

metrics/void-byte-v1/1791300814123050567-72be911d/train.jsonl

@@ -38,3 +38,4 @@
 {"step": 803800, "loss": 0.5099411606788635, "lr": 3e-05, "grad_norm": 0.270241294812033, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117652931556, "tokens_per_second": 152130.01819733303, "data_wait_s": 1.70599996636156e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 803900, "loss": 0.5207116007804871, "lr": 3e-05, "grad_norm": 0.2721884989782707, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117667628222, "tokens_per_second": 153257.60957487288, "data_wait_s": 1.504299962107325e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 804000, "loss": 0.5551673769950867, "lr": 3e-05, "grad_norm": 0.26991183216104825, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117682351354, "tokens_per_second": 153545.42914045625, "data_wait_s": 1.5083000107551925e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 804100, "loss": 0.584531843662262, "lr": 3e-05, "grad_norm": 0.27472927765275135, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117697095316, "tokens_per_second": 152842.4900825801, "data_wait_s": 1.3640999895869754e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791300814.0c9d201b-2dbc-40fb-b739-4ee2350a0f54-7b2tx.840.0.1791300814123050567-72be911d

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2fdbd080835aab5df0e18379dac6fe7a2a51a5f68334b89e8ff41de344166f07
-size 84885
+oid sha256:a3478bd46a196a7fb119aa2b231c35a1356450c4c02d5ebc53f95923a12b3eb3
+size 87003